VR-Zone: FX-9590 is the final legacy of the FX line

[Slomo4shO]

The counter argument is simpler: AMD can't afford to make just a 6 or 8 core CPU, and therefore AMD probably also cannot afford to make a 6 or 8 core APU with a very large graphics component. I think you underestimate how hard it is to change ANYTHING about a design. It's not as simple as "snip that wire, take out that block from the design", it is thousands of manhours of work designing, verifying, constructing, testing etc. for even the most minor alteration to the design.

I am unsure how you draw the conclusion that I am trivializing the R&D required for such a chip. I see it as the logical step from the current offerings of 4 core APUs. Whether or not AMD invests the resources for such an endeavor is yet to be determined.

 

[sushiwarrior]

I am unsure how you draw the conclusion that I am trivializing the R&D required for such a chip. I see it as the logical step from the current offerings of 4 core APUs. Whether or not AMD invests the resources for such an endeavor is yet to be determined.

If they made solely 6/8 core APUs, they would drive the selling price up, and alienate their mainstream market. It is clear that AMD can only afford to do one or the other. Focusing all of their R&D on an expensive 6/8 core + high end graphics APU would be nothing but pure suicide (that would be what, a >400mm^2 chip?).

The logical step is to make a better 4 core APU, with better graphics and better IPC. More cores cost more money.

 

[Exophase]

A derivative of this chip would be a better option than reconfiguring the Jaguar APU for the desktop. The arguement is simple: a 6, 8, or more core APU with a solid GPU that is capable of hybrid crossfire would enhance AMDs position in the mainstream market while encouraging enthusiast to consider AMD for gaming without actually directly competing with Intel on the performance front(AMD as expressed that it has no interest in competing with Intel on performance).

Your idea just doesn't make financial sense. And I'm not sure why you're pushing it so hard. Consider the following:

Your chip with 8 Steamroller cores + HD 7850 level GPU would easily be > 400mm^2. Manufacturing costs don't scale linearly with die area - the bigger you make the die the less efficiently you can fit it on the wafer and the poorer your yield becomes, in addition to other factors. These huge chips are undesirable and you usually only see them in very expensive products. Right from the start it would probably cost more to manufacture than the CPU part and GPU part separately.

Then you have to factor in that there's intrinsically less of a market. AMD's discrete GPUs have wider marketability because they can be paired with any x86 CPU and because they can be put in laptops. This APU would be a bad fit for servers and an impossible fit for laptops.

The console makers are making relatively big APUs (although not nearly as big as what you want, at least not in Sony's case) but it can pay off because of the volume they command, because they'll be shrunk in the future, and because their fixed architecture can better leverage advantages of the APU design.

This chip would need something to address the bandwidth limitation present in current AMD APUs. It would need well over 100GB/s to even be remotely viable. AMD could only accomplish this with some kind of big cache like Intel is doing - which is not what they want to do for Kaveri so would be another huge design expense - or with a really fat GDDR5 interface comparable to GPUs. Which would drive up the price a ton more. It would also evaporate any possible price benefit you got from using unified memory vs the usual DDR3 + GDDR5 split in a CPU + GPU setup. This also means AMD has to roll out a new motherboard socket just for this.

The only reason anyone buys APUs in desktops is to save price. Even IF you'd save money doing this, which I doubt (especially if you need a water cooler), it's a lot less appealing when the base cost is several hundred more than where APUs usually play in. And you end up with something where you can no longer upgrade the CPU or GPU independently. Please explain to me, why do you even want this?

 

[Slomo4shO]

The only reason anyone buys APUs in desktops is to save price. Even IF you'd save money doing this, which I doubt (especially if you need a water cooler), it's a lot less appealing when the base cost is several hundred more than where APUs usually play in. And you end up with something where you can no longer upgrade the CPU or GPU independently. Please explain to me, why do you even want this?

Do you think that your options would increase if AMD left the mid to high end consumer CPU market? I would rather still have an option for a 6+ core AMD APU at the >$150 price point that would still provide some pressure on Intel to improve their lineup. Competition is vital, if AMD refuses to compete with Intel directly then our best hope is indirect competition.

If they made solely 6/8 core APUs

The logical step is to make a better 4 core APU, with better graphics and better IPC. More cores cost more money.

There are limitations to how far you can improve upon a single core. I am unsure where I said that the 2/4 core options would need to be scrapped in lieu of 6/8 core APUs. The current generation of APUs all occupy the sub-$150 market. The future generation of 2/4 core will likely continue to occupy this price block. If AMDs intent is to truly shift to APUs as they have suggested, they will want to have something to offer for the >$150 market unless they intend to completely abandon this market.

 

[Keysplayr]

To you maybe, but for collectors it has a value. Value comes from rarity, and special circumstance. A 486 has almost no collector's value because they were commonplace and everywhere, and never rare. The 9590, if it really is AMD's last performance CPU, is unique in that aspect and fairy rare based on what everyone is saying about the number of them produced.

A 2013 Focus might have more horsepower than the original mustang, but in 50 years a 2013 model Ford Focus is going to be basically worthless while an original '64 Mustang will still be an incredible collector's item.

Eh, you might be right. I'll look you up in sixty.

 

[Exophase]

Do you think that your options would increase if AMD left the mid to high end consumer CPU market? I would rather still have an option for a 6+ core AMD APU at the >$150 price point that would still provide some pressure on Intel to improve their lineup. Competition is vital, if AMD refuses to compete with Intel directly then our best hope is indirect competition.

If AMD can no longer afford to make mid to high end CPUs then that doesn't mean they can afford to make the kinds of APUs you're describing instead! Competition is important but Intel isn't going to care about an expensive novelty APU anymore than they care about FX-9590.

Let me say it again: expensive high end desktop APUs don't exist the sole point of this market is to cut costs.

 

[Slomo4shO]

If AMD can no longer afford to make mid to high end CPUs then that doesn't mean they can afford to make the kinds of APUs you're describing instead! Competition is important but Intel isn't going to care about an expensive novelty APU anymore than they care about FX-9590.

Let me say it again: expensive high end desktop APUs don't exist the sole point of this market is to cut costs.

I am curious to know on what basis are you drawing these conclusions. I am unsure how AMD indicating that they are unable to compete directly with Intel implies that they can't afford the R&D for mid to high end CPUs. Expensive high end desktop APUs don't exist yet. Do you have any data supporting that there isn't a market for them?

 

[frozentundra123456]

I am curious to know on what basis are you drawing these conclusions? I am unsure how AMD indicating that they are unable to compete directly with Intel implies that they can't afford the R&D for mid to high end CPUs. Expensive high end desktop APUs don't exist yet. Do you have any data supporting that there isn't a market for it?

Do you have any data suggesting that there IS a market for one?

Even if they wanted to try to cool basically a 7850 and cpu on top of one another, the performance still will not be there until they solve the bandwidth problem. I just dont see why some posters in these forums are so vested in trying to prove an APU is a great gaming solution for the desktop. There are just too many limitations that are not present with a discrete card for an APU alone to be more than a low end solution.

 

[DeeJayBump]

Slomo-

What you might need to accept is that some of the posters you are arguing with are folks with deep knowledge of both CPU/GPU design and/or fabbing process as well as the current financial standing of both AMD and Intel. The kind of folks that KNOW the intricacies (engineering or otherwise) of doing what you propose and also the kind of folks who keep abreast of AMD's financial filings and conference calls with their BODs, shareholders and investors that substantiate and detail AMD's current financial position.

Just FYI, my $0.02, your posting mileage may vary, etc.

 

[Slomo4shO]

Do you have any data suggesting that there IS a market for one?

I recommended it because I would be interested in such a product. Would make for a great HTPC/Steambox. I may be the only one but, however small it may be, there is a marker for it... 🙂

The limitations that you speak of are current problems. This is agreed. However, the stipulation that these issues can't be resolved...

What you might need to accept is that some of the posters you are arguing with are folks with deep knowledge of both the CPU/GPU-fabbing process as well as the current financial standing of both AMD and Intel. The kind of folks that KNOW the intricacies (engineering or otherwise) of doing what you propose and also the kind of folks who keep abreast of AMD's financial filings and conference calls with their BODs, shareholders and investors that substantiate and detail AMD's current financial position.

I am no engineer and nor do I suggest that I have intricate knowledge of fabrication. Nevertheless, I do like to think that I keep a well enough understanding of the financial standings of both of these companies to refute the claim that AMD is no longer capable of investing in R&D. Technological advancements make possible what was once deemed impossible. The ROI of such project may or may not be worthwhile but concluding something improbable based on current fabrication tech seems a little immature. I would love to hear a real argument on the subject to enlighten myself to the impossibility of such a task but the current refute of product suggestion is nothing more than trollish derailment.

 

[DeeJayBump]

Fair enough. Good luck with that whole banging your head against a wall thing, hope it works out for you.

 

[Slomo4shO]

Fair enough. Good luck with that whole banging your head against a wall thing, hope it works out for you.

Sometimes it takes a thick skull to break through a wall of biases and assumptions :biggrin:

 

[DeeJayBump]

Doesn't look like biases or assumptions to me, but rather economic realities. Let's summarize it shall we?

1. What you propose IS possible, as many in this thread have affirmed.

2. AMD DOES have money for R&D, and they choose to use it wisely on products that have both a significant potential market and a high-probability of financial success.

3. The product(s) you propose have neither a significant potential market nor a reasonable probability of financial success.

4. You continue to refuse to accept these realities and continue to bull-headedly suggest that AMD throw precious funds into R&D and development of products that very few consumers will buy at present and that almost certainly would result in financial disaster to boot.

5. Have I missed any of the pertinent points?

 

[Slomo4shO]

Doesn't look like biases or assumptions to me, but rather economic realities. Let's summarize it shall we?

1. What you propose IS possible, as many in this thread have affirmed.

2. AMD DOES have money for R&D, and they choose to use it wisely on products that have both a significant potential market and a high-probability of financial success.

3. The product(s) you propose have neither a significant potential market nor a reasonable probability of financial success.

4. You continue to refuse to accept these realities and continue to bull-headedly suggest that AMD throw precious funds into R&D and development of products that very few consumers will buy at present and that almost certainly would result in financial disaster to boot.

5. Have I missed any of the pertinent points?

I have a difficulty with #2 and 3 hence why you think I am entangled in #4. There has been no admittance or refutation of what products are down the pipeline other than the fact that the focus is on APU and mobile markets. In conjunction with this fact, the assumption of #3 seems to have no data to either support or refute the market interest of such a product.

If there is a factual basis for these arguments, please feel free to point me in the right direction because I have yet to find anything myself.

 

[mrmt]

Nevertheless, I do like to think that I keep a well enough understanding of theI am no engineer and nor do I suggest that I have intricate knowledge of fabrication. financial standings of both of these companies to refute the claim that AMD is no longer capable of investing in R&D.

This is a red herring. Nobody ever said here that AMD is no longer capable of investing in R&D. What everybody here is saying is that AMD is no longer capable of taking great risks as the one you are suggesting. They have to play small and safe, and what you propose here is to go big and risky.

The kind of product you are would cost many times more than current offers and wouldn't command a correspondent bigger price premium over current parts. With gross margins already on the 40% range, where do you think AMD would stop if they decided to pursue this route?

 

[mrmt]

I have a difficulty with #2 and 3 hence why you think I am entangled in #4. There has been no admittance or refutation of what products are down the pipeline other than the fact that the focus is on APU and mobile markets. In conjunction with this fact, the assumption of #3 seems to have no data to either support or refute the market interest of such a product.

You have evidence of #3, given by AMD itself. They did not pick the kind of product you suggested and placed on their roadmap. It's not a big fork decision like K8 vs Bulldozer that doomed the company, the kind of product you are suggesting is simply a matter of joining up all existing technology and generate a design. And unless you believe that the company is ran by a bunch of incompetents, they would have pursued that route if ROI and TAM were right.

 

[Slomo4shO]

This is a red herring. Nobody ever said here that AMD is no longer capable of investing in R&D. What everybody here is saying is that AMD is no longer capable of taking great risks as the one you are suggesting. They have to play small and safe, and what you propose here is to go big and risky.

The kind of product you are would cost many times more than current offers and wouldn't command a correspondent bigger price premium over current parts. With gross margins already on the 40% range, where do you think AMD would stop if they decided to pursue this route?

Exophase suggested that AMD is no longer capable of manufacturing mid to high end systems. Ergo not capable of investing in R&D for such processors.

I still fail to understand why we continue to discuss current offerings and manufacturing limitations for a product that currently does not exist... but somehow my mention of R&D costs is the red herring?

You have evidence of #3, given by AMD itself. They did not pick the kind of product you suggested and placed on their roadmap.

So you have the details of the complete APU lineup in the upcoming generations?

Either way, good night. We may discuss this further tomorrow 🙂

 

[Idontcare]

I have a difficulty with #2 and 3 hence why you think I am entangled in #4. There has been no admittance or refutation of what products are down the pipeline other than the fact that the focus is on APU and mobile markets. In conjunction with this fact, the assumption of #3 seems to have no data to either support or refute the market interest of such a product.

If there is a factual basis for these arguments, please feel free to point me in the right direction because I have yet to find anything myself.

This is an odd way to approach the world.

Do you walk around assuming Coca-Cola is going to produce a 10L bottle just because 2L and 3L bottles currently exist? All while refusing to consider the practical considerations of why a 10L bottle isn't going to happen?

And in the meantime arguing that unless Coca-Cola themselves come out and tell you, in no uncertain terms, that a 10L bottle isn't happening then you aren't going to listen to anyone else who can just as easily tell you why it isn't on the roadmap?

When you are arguing about something from a position of ignorance, having neither the formal education or firsthand work experience in the relevant engineering areas or business management areas, you are the tail trying to wag the dog.

Arrogance and ignorance are a deadly combination, lots of knowledgeable people here in these forums are more than willing to help the ignorant become educated and more knowledgeable, but not many people are willing to waste their time combating an arrogant individual who seems to just want to argue with anyone and everyone who doesn't share their admittedly ignorant understanding of how things work.

tl;dr - You admit you haven't a clue, but you refuse to accept the ones being offered to you by people who have a clue. To liberally borrow from Mr. Bubbles, thus far you have been adrift in the sheltered harbor of their patience, but you will rapidly lose their attention and opportunity to get yourself a clue if you persist in this manner of argumentation from a position of admitted ignorance.

 

[bononos]

Originally Intel actually gave AMD the designs, they had to because AMD was actually a foundry for Intel. You have to give the foundry the designs (the masks). The foundry doesn't have to analyze them, but they easily and readily can if they want to.

Where things got problematic was when Intel decided they didn't want AMD as a foundry any more, and did not extend their contract to the newer processors that Intel was coming out with.

......

But how did AMD become a foundry for Intel? I think IBM wanted more sources for its cpus and forced Intel to license NEC, Siemens, AMD to make 8088s. And it was a good decision not to hand over a monopoly to Intel like they did to Microsoft.

 

[Idontcare]

But how did AMD become a foundry for Intel? I think IBM wanted more sources for its cpus and forced Intel to license NEC, Siemens, AMD to make 8088s. And it was a good decision not to hand over a monopoly to Intel like they did to Microsoft.

It was at IBM's insistence, of course.

At the time it wasn't about monopoly concerns, rather the opposite. Intel was a rather small-time MPU logic manufacturer.

It was simply supply-chain risk-management from IBM's POV.

IBM's requirement that Intel provide multiple production sources had nothing to do with licensing x86 to multiple 3rd parties. That was actually Intel's solution to IBM's requirement.

What IBM didn't want was a business line that was physically dependent on a single fab or geographic region for the chips. Intel was told to establish more fabs with greater geographic separation. If a tornado, hurricane, flood, or earthquake disrupted production at one fab then at least IBM's product line wouldn't be in jeopardy.

IBM was fine with Intel investing into more production facilities of their own. But at the time even Intel wasn't so sure that x86 was going to be the mega-CPU that it eventually became. In other words Intel wasn't so excited about plunking down the capex to build more fabs to reduce IBM's supply risks.

So Intel took the measure of reducing its own "demand" risk for capex by licensing x86 as a means of securing foundry services from existing established fabs that were already spread around the globe.

IBM was fine with Intel's proposed solution, it served the purpose of risk mitigation in the supply-chain on multiple levels. That it came back to bite both of them in the rear-end is just irony.

And this is also why IBM was not concerned about Microsoft, there was no supplier-chain risk to mitigate in that part of the business. No natural catastrophe to consider as putting your own product line in peril, so no need to insist on multiple-sources for the supplies.

That Microsoft would go on to become a monopoly was not readily obvious at the time. It isn't practical to treat every nascent business venture as being the progenitor of the next decade's monopoly. If businesses interacted that way then nothing would ever get done.

 

[Charles Kozierok]

IBM was fine with Intel investing into more production facilities of their own. But at the time even Intel wasn't so sure that x86 was going to be the mega-CPU that it eventually became. In other words Intel wasn't so excited about plunking down the capex to build more fabs to reduce IBM's supply risks.

Even more interesting is that Intel was at first skeptical about the idea of selling CPUs entirely. They began in memory, and for a while their view was that this was a more lucrative business because a computer needs many memory chips but only one CPU. They only agreed to design the 8008 (predecessor of the 8080 and then 8086/8088) because a customer gave them a contract for it.. and then they ended up owning the chip when that company decided not to take delivery.

Beyond not worrying about monopolies, IBM and Microsoft were actually development partners for a while (OS/2) before Gates and Co. decided to push Windows instead.

 

[mrmt]

Exophase suggested that AMD is no longer capable of manufacturing mid to high end systems. Ergo not capable of investing in R&D for such processor

No, he didn't. Despite the cuts AMD still has a very respectable R&D muscle yes, they would be able to develop the kind of product you are suggesting. There is no new tech anyway, just put existing parts together. What Exophase and I suggested is that AMD no longer has the capabilities to bring to the market such processor.

AMD financial resources are already near minimum levels and they simply cannot afford to risk money in a risky project like the one you are suggesting. One wrong move and they are done for. And if AMD cannot bear the risk, AMD can't build the project.

There is a big difference between seeing an immensely flawed product/concept like Bulldozer, Netburst, Zune and seeing a flaw in the fine tuning of a product line like the one you are suggesting. It's easy to see a big flaw in a concept or in execution like the three examples I mentioned. You can safely bet that this was seen internally but management for some reason (some rational like TTM or strategic calculus, others not so rational like pride, hubris or sheer stupidity) proceeded with the venture.

In the same tune, it's easy to see that APU is a dog for margins, but once AMD management gave the marketing guys Bulldozer and decided to go APU you can bet that there was a marketing team full of dedicated people running every simulation possible and gathering all the marketing data they could in order to make the most money out of the thing, and if that team decided that it wasn't worth to go to huge APUs, then you can almost always bet that the right answer is to not to go huge APU.

Just look at the last two paragraphs. The scope in the former case is huge, and so is the number of variables that can impact the thing. You from your armchair have a very good chance to hit some of the most significant variables (as they are out in the environment) and you may be right in your judgement. For fine tuning, the scope is far smaller, but you wouldn't have most of the critical parameters like R&D costs, projected performance of the product, TAM, target ROI, COGS and OEM input because they are internal.

You are essentially saying that you can do better than AMD marketing guys just because *you* are interested in such a product.It's easier to guess management than to guess the guy running the model.

 

[Slomo4shO]

No, he didn't.

Oh?

If AMD can no longer afford to make mid to high end CPUs then that doesn't mean they can afford to make the kinds of APUs you're describing instead!

Seems you like to only read between the lines...

This is an odd way to approach the world.

You admit you haven't a clue, but you refuse to accept the ones being offered to you by people who have a clue. To liberally borrow from Mr. Bubbles, thus far you have been adrift in the sheltered harbor of their patience, but you will rapidly lose their attention and opportunity to get yourself a clue if you persist in this manner of argumentation from a position of admitted ignorance.

My initial statement was

An 8 core APU with an integrated GPU as powerful as the HD 7850 (PS4 equivalent graphics power) for around $200-220 would definitely be something worth consideration. Especially if the integrated GPU can be run in hybrid crossfire with any of the HD 7000 series or newer GPUs.

and I was meet with nothing more than a dogmatic response of the impossibility of such a product. My argument was never that AMD will be producing this chip or the is capable of producing this chip with the current technology but, apparently, that is the only thing anyone is willing to discuss on these forums.

Goals enable possibilities, not the other way around. Apparently the individuals "who have a clue" on these forums also wear blinders.

 

[mrmt]

Oh?



Seems you like to only read between the lines...

AMD financial resources are already near minimum levels and they simply cannot afford to risk money in a risky project like the one you are suggesting. One wrong move and they are done for. And if AMD cannot bear the risk, AMD can't build the project.

It's not a "do I have money or not" question, is a "can I bear the risk even if this has a significant probability of success?".

 

[Slomo4shO]

It's not a "do I have money or not" question, is a "can I bear the risk even if this has a significant probability of success?".

You continue to read between the lines and draw you own conclusions since I addressed neither of these questions in any of my posts.

I am curious to know on what basis are you drawing these conclusions. I am unsure how AMD indicating that they are unable to compete directly with Intel implies that they can't afford the R&D for mid to high end CPUs. Expensive high end desktop APUs don't exist yet.

This is the closest reference to these questions but it was a rebuttal to a claim and not a direct analysis of either of these questions. Feel free to continue to draw your own conclusions that actually have no relevance to any of my posts :whiste: